Patent Valuation Jonathan Putnam Page 1

Patent Valuation Jonathan Putnam Page 1 The valuation of patent rights sounds like a simple enough concept. It is true that agents routinely appraise and trade individual patents. But small-sample methods (generally derived from basic accounting and finance) are often crude, and their results may bear little relationship to economic fundamentals, especially in litigation. On the other hand, large-sample methods usually lack much invention-specific data on which to condition value estimates. Regardless of sample size, proper valuation methods require both conceptual delineation and empirical ingenuity. Concepts. Legally, a patent is the right to exclude others from making, using or selling an invention. In economic terms, that right is an asset, yielding a non-negative returns stream while it is enforceable. Because the right is a private means (increased exclusivity) to a public end (increased productivity), a patent’s private value only partially conveys its market significance. Unlike most property rights, patents do not comprise the affirmative right to use the invention. Absent the right to use, patents may generate private value only when combined with complementary assets, such as a license under other patents. Contracting problems (e.g. asymmetric information) may strongly influence value. A patent may generate private returns apart from the right to exclude rivals. The patentee may use it: to monitor employee performance; to signal otherwise unobservable quality to prospective financiers; to enhance reputation; to signal a willingness to litigate; or to reduce the costs of settlement in the event that litigation occurs (“defensive patenting”). In large samples, it is usually impossible either to observe the magnitude and timing of these sources of value, or to decompose them. Patent Valuation Jonathan Putnam Page 2 Patents also impose unobservable private costs on the patentee. Chiefly, the inventor must disclose the means for reproducing the invention. Disclosure reduces the cost to rivals of reproducing the invention (static spillover) and conducting R&D (dynamic spillover). Apart from reducing the incentive to invent, these private costs imply social benefits not captured by the patentee. Cross-sectionally, patents are usually modeled as having a one-dimensional “quality” (which is either synonymous with, or a monotone function of, the patent’s value). More precisely, a patent’s private value depends significantly on the exclusivity conferred by its claims, but its uncaptured social value depends significantly on the scope of its disclosure (which must be at least as broad as the claims). For various reasons, including rival use of the patentee’s disclosure to develop competing innovations (“creative destruction”), the social and private values of a patent may diverge. Thus, it is theoretically preferable, but empirically much less tractable, to model patents as having two-dimensional “quality.” Over time, because of ongoing research by the patentee and his rivals, the private returns to patent protection may fluctuate sharply up or down, in response to complementary or competitive discoveries. The variance is likely to be larger in a patent’s early years. Stylized facts. The following stylized facts bear on the calculation of aggregate private patent values: 1. Whether aggregated by firm, industry or country, patent counts do not vary much from one period to the next. Patent Valuation Jonathan Putnam Page 3 2. The distribution of patent values is skewed. 3. Social and private patent values are imperfectly correlated. 4. Ex ante and ex post values are imperfectly correlated. 5. Most patents are not traded. 6. Samples are selected (not all innovations are patented; not all applications are filed in any single country; not all applications are granted). Related research Proceeding in the direction of generally increasing complexity and structure, the following categories describe large-sample models that economists have developed to value patent rights. Lanjouw, Pakes and Putnam (1998) surveys recent papers. Patent counts. A variety of models employ simple patent counts to indicate the value of patent rights. Strictly speaking, patent counts indicate quantities, rather than values. Under certain assumptions, relative quantities may be proportional to relative values. For example, if two patent samples are drawn from the same value distribution, then the ratio of quantities is an efficient estimator of the ratio of values. Griliches (1990) reviews a large number of studies that, implicitly or explicitly, rely on this assumption. Griliches’ view of “patent [counts] as economic indicators” is not encouraging (“The food here is terrible.” “Yes, and the portions are so small.”). Stylized facts #1 and #2 combine to thwart inference. A firm facing a fixed budget constraint may patent its best N inventions, which implies little intertemporal variation in patent counts even if their realized quality varies markedly. Thus, patent counts are a biased measure of value. Because R&D outcomes are highly variable and skewed, patent Patent Valuation Jonathan Putnam Page 4 counts are an imprecise measure of value. For these reasons, the assumption that patent samples are drawn from the same distribution is difficult to test, and often false. (On the other hand, fixed budget constraints for R&D and patenting imply that patent counts may proxy for the value of R&D inputs. Hausman, Hall and Griliches (1986) model the lag relationship between patent counts and R&D, and find an approximately contemporaneous relationship.) One may compute implied patent values by associating patent counts with other observable aggregates. On the macro level, McCalman (2005) employs the structural imitation model of Eaton and Kortum (1996) to determine international “trade” in patents. He estimates that the worldwide value of patent applications filed by U.S. inventors in 1988 was about $12.4 billion ($163,700 per application). The estimates for four other large patenting countries vary: France, $147,200; Germany, $82,200; U.K., $53,100; Japan, $47,700. At the firm level, Pakes (1985) constructs a time series model of patent applications, R&D and the stock market rate of return. Controlling for R&D expenditures, an unanticipated patent application implies an $800,000 increase in market capitalization. (This relatively high value also reflects investors’ revised expectations of research success, and the selection of publicly traded patentees (which are larger and more successful than average). Patent citations (weighted patent counts). Patent examiners cite prior patents when they decide whether to grant a patent application. Analysts count these citations to Patent Valuation Jonathan Putnam Page 5 indicate the value of the cited patent. Patent counts are then weighted by the number of citations. A recent book-length treatment is Jaffe and Trajtenberg (eds.) (2002). This branch of the literature divides in two: estimates of the relationship between citations and patent value; and studies that assume that relationship. In the former category, Trajtenberg’s (1990) pioneering study showed that citation-weighted patent counts perform better than unweighted counts in explaining aggregate patent value (see Harhoff et al. 2003). However, this and subsequent studies found that citations tend to indicate the social value of the patent, rather than the purely private value (stylized fact #3). Private value is better captured by “self-citations” from the patentee’s own later inventions. Hall et al. (2005) show that weighted patent counts are associated with—and predict—higher stock market returns. Assuming that citations proxy for value, Henderson, Jaffe and Trajtenberg (1998) examine the contribution of university patenting to commercial technology; Trajtenberg, Henderson and Jaffe (1997) find that the “basicness” of university patents relative to corporate patents has narrowed over time. Jaffe, Trajtenberg and Henderson (1993) model the spatial distribution of dynamic spillovers. Other indicator-based methods. Lanjouw and Schankerman (2004) construct a composite index of patent quality using several indicators (forward- and backward- citations, number of claims, and number of filing countries). This combination of ex ante and ex post measures (stylized fact #4) efficiently aggregates informationally distinct components of patent value. The composite also explains related ex post decisions (e.g., Patent Valuation Jonathan Putnam Page 6 patent renewal and litigation); forward citations (an ex post measure) demonstrate the greatest explanatory power. Structural models: patent renewals and patent applications. Although most patents are not traded (stylized fact #5), patent office rules effectively require patentees to make optimal investments to create and maintain patent rights. These investments reveal information about the expected value of the asset. The information is censored, however, because (conditional on choosing to invest) patentees make the same investment regardless of the expected value. Structural econometric models identify the underlying value distribution. Most countries require that a patentee pay an increasing fee to keep a patent right in force. Beginning with Pakes and Schankerman (1984), so-called patent renewal models exploit the optimal stopping problem implicit in the annual investment decision. The ex post value distribution is identified from the shares of an annual cohort that are renewed each subsequent year when patentees confront

Load more